Table 2.

Comparison of SAMGSR with other feature selection algorithms.

Method	Subtype	Training set				TCGA RNA-seq
		Error (%)	GBS	BCM	AUPR	Error (%)	GBS	BCM	AUPR
SAMGSR + SVM	AC (119)	0	0.043	0.810	0.996	42.9	0.350	0.609	0.630
	SCC (26)	2.36	0.062	0.802	0.992	32.7	0.256	0.589	0.583
Lasso	AC (81)	0	1.14 × 10−4	0.990	0.996	35.7	0.357	0.5	0.624
	SCC (33)	0	<10−4	0.993	0.992	29.1	0.291	0.5	0.565
Penalized	AC (528)	0	0.003	0.951	0.996	37.1	0.318	0.524	0.615
SVM (SCAD)	SCC (63)	0	<10−4	0.999	0.959	27.3	0.273	0.531	0.654
DEGs + SVM	AC (145)	0	0.042	0.810	0.996	51.9	0.465	0.562	0.638
	SCC (46)	0	0.046	0.803	0.992	29.1	0.287	0.501	0.632
Radviz + SVM	AC (9)	22.83	0.166	0.559	0.734	37.1	0.363	0.493	0.541
	SCC (8)	4.76	0.076	0.774	0.934	30.9	0.293	0.493	0.536